A model for predicting bacteremia in patients with community-acquired pneumococcal pneumonia: a retrospective observational study

Yasuyoshi Washio, Akihiro Ito, Shogo Kumagai, Tadashi Ishida, Akio Yamazaki

Pneumococcal pneumonia causes high morbidity and mortality among adults. This study aimed to identify risk factors for bacteremic pneumococcal pneumonia, and to construct a prediction model for the development of bacteremia in patients with community-acquired pneumococcal pneumonia. We retrospectively analyzed data from patients hospitalized with community-acquired pneumococcal pneumonia between April 2007 and August 2015. Logistic regression models were applied to detect risk factors for pneumococcal bacteremia, and a receiver operating characteristic curve was used to devise a prediction model. Based on the results of sputum cultures, urine antigen tests, and/or blood cultures, 389 patients were diagnosed with pneumococcal pneumonia, 46 of whom had bacteremia. In the multivariate analysis, age < 65 years, serum albumin level < 3.0 g/dL, need for intensive respiratory or vasopressor support (IRVS), and C-reactive protein level > 20 mg/dL were identified as independent risk factors for the development of pneumococcal bacteremia. The bacteremia prediction score based on receiver operating characteristic curve analysis had a sensitivity of 0.74 and a specificity of 0.78 in patients with two risk factors. The area under the receiver operating characteristic curve was 0.77 (95% confidence interval (CI), 0.70-0.85). Age < 65 years, hypoalbuminemia, IRVS, and high C-reactive protein level on admission are independent risk factors for the development of bacteremia in patients with community-acquired pneumococcal pneumonia. A prediction model based on these four risk factors could help to identify patients with community-acquired pneumococcal pneumonia at high risk of developing bacteremia; this can be used to guide antibiotic choices. UMIN-CTR UMIN 000004353 . Registered 7 October 2010. Retrospectively registered.